Search

Assistive aerial teleoperation in unknown environments can be challenging due to the limited sensing range of cameras and the insufficient human–computer interaction. Obstacles occluding the field of view (FOV) of camera reduce the visibility of unknown environments, and it coupled with insufficient pilot instructions directly affect human intention recognition. To address this issue, the planned trajectory generated by assistive teleoperation should align with human intention while considering perception awareness (PA) of the FOV. However, most of the existing works considering PA solely increase the FOV visibility while neglect the human intention constraints. As a result, the PA constraints can cause the heading of the quadrotor to deviate from human expectation, requiring more human–computer interactions to correct it. In this paper, we propose a human-guided motion planner with perception awareness to solve these issues. Specifically, in the path-finding stage, we introduce a spatio-temporal path similarity metric method. This method, combined with motion primitive propagation, finds an initial path that is collision-free and roughly aligned with human intention. In the subsequent trajectory optimization stage, we jointly consider PA with respect to occlusion and human intention constraints, effectively optimizing FOV visibility without deviating the heading of quadrotor from human intention. Extensive experiments and benchmark comparisons demonstrate the effectiveness of our proposed method in solving assistive aerial teleoperation for even novice operators. [ABSTRACT FROM AUTHOR]

Direct visual odometry (DVO) is an important vision task which aims to obtain the camera motion via minimizing the photometric error across the different correlated images. However, the previous research on DVO rarely considered the motion bias and only calculated using single direction, therefore potentially ignoring useful information compared with leveraging diverse directions. We assume that jointly considering forward and backward calculation can improve the accuracy of pose estimation. To verify our assumption and solid this contribution, in this paper, we test various combination of direct dense methods, including different error metrics, e.g., (intensity, gradient magnitude), alignment strategies (Forward-Compositional, Inverse-Compositional), and calculation directions (forward, backward, and bi-direction). We further study the issue of motion bias in RGB-D visual odometry and propose four strategy options to improve pose estimation accuracy, e.g., joint bi-direction estimation; two stage bi-direction estimation; transform average with weights; and transform fusion with covariance. We demonstrate the effectiveness and efficiency of our proposed algorithms across a range of popular datasets, e.g., TUM RGB-D and ICL-NUIM, in which we achieve an impressive performance through comparing with state of the art methods and provide benefits for existing RGB-D visual odometry and visual SLAM systems. [ABSTRACT FROM AUTHOR]